The design of the GraphBLAS Template Library separates graph algorithm development from performance tuning for heterogeneous high-performance computing architectures.
In this paper we present our initial implementation of GraphBLAS primitives for graphics processing unit (GPU) systems called GraphBLAS Template Library (GBTL).
Authors: Peter Zhang (Indiana University), Eric Holk (Indiana University), John Matty, Samantha Misurda, Marcin Zalewski (Indiana University), Jonathan Chu, Scott McMillan, Andrew Lumsdaine (Indiana University)
Presented at the 2015 Supercomputing Conference, this paper shows that dynamic parallelism enables relatively high-performance graph algorithms for GPUs.
Authors: Jeremy Kepner (No Affiliation), Henning Meyerhenke (Karlsruhe Institute of Technology), Peter Aaltonen (Indiana University), David Bader (No Affiliation), Aydın Buluç (Lawrence Berkeley National Laboratory), Franz Franchetti (Carnegie Mellon University), John Gilbert (No Affiliation), Dylan Hutchison (University of Washington), Manoj Kumar (IBM), Andrew Lumsdaine (Indiana University)
This paper introduces the mathematics of the GraphBLAS, which is being developed to bring matrix-based graph algorithms to the broadest possible audience.